Some topics are off limits inside popular Chinese-made free AI, researchers find, tell me about this news today
02-10-2026
An investigation published by CBS News focused on Alibaba’s popular, free, and open-weight artificial intelligence model, Qwen. Researchers and testing revealed significant political guardrails, censorship, and systemic alignment with official Chinese state viewpoints:
CBS NewsOff-Limits and Sanitized Topics: When prompted about sensitive historical or political matterssuch as the 2019 Hong Kong protests, the 1989 Tiananmen Square crackdown, human rights records, or top leadershipthe model either avoids directly answering, deflects, or aligns strictly with the official narrative of the Chinese government. For instance, tests showed the model failing to acknowledge the violent suppression of demonstrations or the existence of certain state controversies.
CBS NewsGlobal Reach vs. State Censorship: While censorship inside Chinese tech services has long been common under domestic Cyberspace Administration regulations, researchers expressed concern because Qwen has gained massive traction outside of China. Its high performance benchmarks and free/open availability have led developers, international companies, and everyday users worldwide to integrate it into their own applications.
CBS NewsRisk of Entrenched Bias: Researchers highlighted the risk that foreign developers and consumers adopting Chinese open-weights could inadvertently embed Beijing-aligned censorship filters and state-approved biases into global digital infrastructure and software workflows.
CBS News